What is thrashing? How might it be detected? How might one recover from it once detected?

Thrashing is when you're trying to run java 'hello world', and any other process on a system. The memory requirements of the currently running processes are so large that the working sets of the executing processes cannot be contained in physical memory. So one process tries to run, gets a page fault, retrieves the page from swap, boots out another processes' frame doing that. Then it context switches and the other process tries to run, but it's memory has been booted, so it page faults to swap, etc. The time is spent swapping frames rather than getting anything done.

It could be detected by monitoring swapping levels compared to real cpu utilization. If swapping is huge, and utilization is down, thrashing is occuring. Suspending processes can alleviate thrashing; hopefully some processes will go below the thrashing threshold, run, terminate, then free up memory for other processes. Installing more ram helps, and not running java hello world helps.


Thrashing occurs when too many processes are run on a processor at a given time. What occurs is the the number of page faults increase dramatically and the virtual memory subsystem is constantly paging pages in and out of memory. This occurs when the working set of all processes is larger than the amount of RAM available on a system.

This can be detected by monitoring the page fault frequency and CPU utilisation. If increasing the number of processes results in increasing page fault rate and decreasing CPU utilisation, then the system is thrashing.

To recover from this condition the number of processes currently in the running/ready queue must be reduced. This can be accomplised by suspending processes (pushing them onto the sleeping queue), so that pressure on physical memory is reduced (suspended processes eventually get swapped out), and thrashing subsides.

1 comments:

Feel free to contact the admin for any suggestions and help.